This clinical research is a retrospective, feature comparison reader study with an enriched sample of 600 (100 per radiologist) breast screening or diagnostic cases which were selected from the library of mammograms collected under Fujifilm protocol FMSU2013-004A (all subjects previously provided written informed consent agreeing their image data and supporting documentation could be used for future research and investigations).
Six MQSA-qualified and board-certified mammographers representing various experience levels, will be shown DBT images of a breast with the modified processing side-by-side the same breast with the original processing on the ASPIRE Bellus II workstation.. One breast per subject (two views per breast; CC and MLO) will be evaluated by the readers. Readers will be blinded to the image reconstruction and presentation processing methods. Each reader will be given a different sample of cases to review. Approximately 600 total cases (100 cases per reader) will compose the sample. The order of the subjects and location (left or right review workstation high resolution monitor) of the images with modified processing will be randomized for each reader.
The primary objective is the readers' assessment for each of the 7 general mammographic features. The null hypothesis is that the pooled proportion of cases judged as non-inferior with modified image processing is \<0.50, versus the alternative hypothesis that the pooled proportion is \>0.50. If the 95% CI does not contain values \<0.50, then the null hypothesis will be rejected. These hypotheses will be evaluated for each of the 7 general mammographic features. If all 7 null hypotheses are rejected, then it will be concluded that the modified imaging features are non-inferior to the original imaging
